Fractional factorial design¶

An experimental design using a structured subset of full-factor combinations to estimate selected effects with fewer runs at the cost of aliasing.

Version: v2 · 2026-09-19 · History
Domain-specific #: 4600
Domain group: Formal Sciences
Origin domain: Experimental Design & Statistics
Subdomain: Design of Experiments → Experimental Design & Statistics

Core Idea¶

Regular and nonregular fractions, resolution, generators and foldover plans determine which effects are confounded; sparsity and hierarchy assumptions justify the economy. Defining contrasts select a fraction of the treatment cube, causing columns for certain effects to coincide while preserving estimability of prioritized low-order effects. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

The load-bearing residual is not the broad topic of design of experiments. It is the domain-specific identity determined by the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model are explicit.

How would you explain it like I'm…

Try Only Some Recipes

Say you want to find the best cookie and you can change the sugar, the butter, and the oven heat. Trying every mix takes lots of baking. A fractional factorial design is a clever plan to bake only some of the mixes. You still learn what matters most, but a few smaller effects get mixed up together, and you choose ahead of time which ones.

Testing Some Mixes, Not All

In an experiment with many factors, each set at a few levels, testing every combination takes a huge number of runs. A fractional factorial design tests only a cleverly chosen part of all the combinations. Because you skip some, certain effects become impossible to tell apart; they are mixed together, or confounded. The design is arranged so the most important simple effects, like each factor on its own, can still be measured. This works because usually only a few factors matter a lot, and simple effects tend to matter more than complicated combinations.

Aliased Partial Factorial Experiments

A fractional factorial design runs only a planned fraction of all factor-level combinations in a full factorial experiment. The fraction is chosen using generators, which define some factor settings from others, giving a defining relation. This makes certain effects aliased (confounded): their columns in the design are identical, so the data can't separate them. The design's resolution summarizes how low-order the aliased effects are; higher resolution keeps main effects clear of two-factor interactions. The economy is justified by effect sparsity (few effects matter) and hierarchy (main effects matter more than interactions). Foldover runs can later add the mirror fraction to break specific aliases.

A fractional factorial design selects a subset of the treatment combinations of a full factorial, typically a 2^(k−p) fraction of a 2^k design. Defining contrasts, specified through generators and the resulting defining relation, determine the fraction; as a consequence, the model columns for certain effects coincide, producing an alias structure. Resolution classifies that structure: in a resolution III design main effects are aliased with two-factor interactions, while resolution IV and V protect progressively more low-order effects. Regular fractions have complete aliasing defined by the defining relation; nonregular fractions have partial aliasing. Foldover plans add complementary runs to de-alias chosen effects. The economy rests on sparsity and hierarchy assumptions. A properly specified design states factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate, and analysis model.

Scope of Application¶

Fractional factorial design belongs to design of experiments and is useful where the analyst can specify the typed design of experiments carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, then evaluate the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model are explicit. The scope is broad within that domain but bounded by the need for the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model are explicit.

Clarity¶

The abstraction clarifies a crowded vocabulary by making the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model are explicit the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test.

Manages Complexity¶

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Fractional factorial design. Fractional factorial design compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning¶

Identify the carrier. State what the elements, states, objects, or observations are: the typed design of experiments carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets. Reject examples whose alleged carrier belongs to a different problem.
Lock the constitutive rule. Express the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model are explicit independently of one notation or implementation.

Knowledge Transfer¶

Knowledge transfers strongly among subfields of design of experiments because they reuse the typed design of experiments carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, Defining contrasts select a fraction of the treatment cube, causing columns for certain effects to coincide while preserving estimability of prioritized low-order effects., and type the carrier, state every parameter and convention in the definition, test that the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions¶

Current abstraction Fractional factorial design Domain-specific

Parents (1) — outgoing typed relations

Fractional factorial design is a kind of Experimental Design Prime

It remains its own entry because its identity is fixed by the factors and levels, full design and fraction size, generators and defining relation, randomization and replication, resolution and alias structure, estimable effects, error estimate and analysis model.

Hierarchy paths (2) — routes to 1 parentless root

Fractional factorial design → is a kind of Experimental Design → includes, typical Control Sample → presupposes Comparison

Show alternative path (1)

Neighborhood in Abstraction Space¶

Fractional factorial design sits in a crowded region of the domain-specific corpus (28^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Sampling, Monte Carlo & Estimation (23 abstractions)

Nearest neighbors

Main effect — 0.93
Analytic and enumerative statistical studies — 0.89
Demand characteristics — 0.88
Regression analysis — 0.88
Two-way analysis of variance — 0.88

Computed from structural-signature embeddings · 2026-10-25